🔬 Export AI agent conversation logs from Postgres to Google Sheets

⚡ 1,071 views · 🔬 Document Extraction & Analysis

Description

Overview

This n8n workflow retrieves AI agent chat memory logs stored in Postgres and pushes them to Google Sheets, creating one sheet per session. It’s useful for teams building chat-based products or agents and needing to review or analyze session logs in a collaborative format.

Who is it for

• Anyone with an AI Agent in Production storing the conversation logs in Postgres (or Supabase) who wants to see transcript and have control
• Product teams building AI agents or assistants.
• Teams that want to centralize conversation history for analysis or support.
• Anyone managing AI chat memory and needing to explore it in a spreadsheet.

Prerequisites

• A Postgres database with a n8n_chat_histories table with an AI Agent connected to it. If you need an example, you can follow this tutorial
• Once done, you need to run the Postgresql query to add the created_at column (see Setup > Add a datetime column)
• Google Sheets access and OAuth credentials connected to n8n.
• A Google Sheets document set up as a template (see below).

Google Sheets Template

This workflow expects a Google Sheets file where each session will be stored in its own tab. A basic tab layout is duplicated and renamed with the session ID. 👉 Use this template as a starting point Note: You can hide the template after the first tabs have been created

How it works

1. Trigger The workflow can be launched manually or on a schedule (e.g. daily at noon).

2. Retrieve sessions Runs a SQL query to get distinct session_id values from the n8n_chat_histories table.

3. Loop over sessions For each session:

   • Clears the corresponding sheet (if it exists).
   • Duplicates the template tab.
   • Renames it with the current session_id.

4. Fetch messages Selects all messages linked to the session from Postgres.

5. Append to sheet Adds each message to the Google Sheet with columns:

   • Who: speaker role (user, assistant, etc.)
   • Message: text content
   • Date: timestamp from created_at, formatted yyyy-MM-dd hh:mm:ss

Notes

• The sheet is cleared and rebuilt each run to ensure logs are up-to-date.
• If a sheet for a session doesn’t exist, it will be created by duplicating the first tab (template)
• You can group sessions under a persistent ID (like user_id) by overriding session_id in your memory config.
• Works perfectly with Supabase by using PG credentials from the connection pooler.

👉 If you’re looking for a solution to better visualize and analyse conversations, reach out to us!

🔗 Nodes Used

Google Sheets, HTTP Request, Postgres, Schedule Trigger

📥 Import

Download workflow.json and import into n8n: Workflow menu → Import from File

📖 Importing guide · 🔑 Credential setup